In the present paper, a method is proposed to improve the performance of Artificial Neural Network (ANN) based algorithms for the retrieval of oceanic constituents in Case Ⅱ waters. The ANN-based algorithms have been...In the present paper, a method is proposed to improve the performance of Artificial Neural Network (ANN) based algorithms for the retrieval of oceanic constituents in Case Ⅱ waters. The ANN-based algorithms have been developed based on a constraint condition, which represents, to a certain degree, the correlation between suspended particulate matter (SPM) and pigment (CHL), coloured dissolved organic matter (CDOM) and CHL, as well as CDOM and SPM, found in Case Ⅱ waters. Compared with the ANN-based algorithm developed without a constraint condition, the performance of ANN-based algorithms developed with a constraint conditions is much better for the retrieval of CHL and CDOM, especially in the case of high noise levels; however, there is not significant improvement for the retrieval of SPM.展开更多
文摘In the present paper, a method is proposed to improve the performance of Artificial Neural Network (ANN) based algorithms for the retrieval of oceanic constituents in Case Ⅱ waters. The ANN-based algorithms have been developed based on a constraint condition, which represents, to a certain degree, the correlation between suspended particulate matter (SPM) and pigment (CHL), coloured dissolved organic matter (CDOM) and CHL, as well as CDOM and SPM, found in Case Ⅱ waters. Compared with the ANN-based algorithm developed without a constraint condition, the performance of ANN-based algorithms developed with a constraint conditions is much better for the retrieval of CHL and CDOM, especially in the case of high noise levels; however, there is not significant improvement for the retrieval of SPM.
